China’s AI hardware sector is pushing ahead with aggressive fundraising and listing plans, even as global chip sentiment wavers. Nvidia shares slumped more than 6% between Tuesday and Thursday under pressure from SoftBank’s sell-off and bearish comments from well-known short seller Michael Burry.
Yet on the same day, two companies often dubbed “China’s Nvidia” reported major progress in their IPO processes, signaling growing investor confidence in domestic GPU developers amid escalating U.S.–China tech tension.
The China Securities Regulatory Commission announced that the IPO registration of high-performance GPU developer MetaX for Shanghai’s STAR Market had been approved. The Beijing-based company, which positions itself as a full-stack provider of general-purpose GPUs for heterogeneous computing, designs chips for AI inference, cloud computing, autonomous driving, digital twins, the metaverse, and smart city applications.
MetaX's key offerings include the XiSi N-series, targeting AI inference, and the XiYun C-series, designed for AI training, inference, and general-purpose computing—both directly benchmarked against Nvidia’s products. With U.S. export controls forcing Nvidia and AMD to scale back their China-market GPUs, MetaX has emerged as a strategic contender in the drive for domestic chip substitution.
MetaX filed its STAR Market prospectus on June 30, seeking to raise 3.904 billion yuan to fund next-generation GPU R&D and industrialization efforts. The company remains unprofitable: its prospectus shows losses of more than 1.4 billion yuan in 2024, and over 200 million yuan in the first quarter of 2025. MetaX estimates its losses narrowed to between 300 million and 380 million yuan in the first three quarters of this year. Still, revenue is accelerating sharply, with order value by the third quarter already more than doubling last year’s full-year revenue. The company has previously said it could reach break-even as early as 2026.
Later that evening, fellow GPU developer Moore Threads released its STAR Market prospectus, formally initiating its listing process after years of being informally described as “China’s first GPU stock.” Founded in 2020 and focused on full-featured GPUs, Moore Threads aims to build globally competitive acceleration-computing infrastructure for industries undergoing digital and AI upgrades.
According to the listing announcement, Moore Threads will trade under the ticker 688795. Price inquiries begin November 19, with subscription slated for November 24. The company plans to issue 70 million shares, representing 14.89% of its post-IPO capital. Proceeds will support development of next-generation AI training-and-inference integrated chips, graphics processors, and AI SoCs.
Like MetaX, Moore Threads has yet to reach profitability. It expects 2024 losses to exceed 1.6 billion yuan, with a first-half loss of 270 million yuan. Cumulative unrecouped losses reached about 1.6 billion yuan as of June 30. But revenue growth has been striking: from 2022 to 2024, the firm achieved a compound annual growth rate of 208.44%.
First-half revenue this year surpassed last year’s full-year total by 60%, with full-year growth projected between 178% and 242%. Still, the company noted that absolute revenue remains modest relative to its heavy investment, with 4.3 billion yuanpoured into R&D over the past three and a half years versus about 1.3 billion yuan in revenue.
Moore Threads sees strong tailwinds ahead as large language models, embodied intelligence, and autonomous driving continue to expand China’s demand for computing power. The company projects continued revenue growth and suggests it could achieve consolidated profitability as soon as 2027.
The two startups’ smooth progress reflects ongoing reforms to the STAR Market’s IPO mechanism. MetaX moved from application submission to registration approval in 116 days, while Moore Threads reached the same milestone in just 88 days, underscoring regulators’ push to support strategic technologies.
China’s tech giants are also doubling down on chip self-reliance. At the Baidu World 2025 Conference on November 13, Baidu reaffirmed its “all-in AI” strategy. Beyond unveiling new AI products, Baidu announced that its self-developed Kunlun Chip M100 will launch in early 2026, followed by the M300 in early 2027. The M100 is optimized for large-scale inference with an emphasis on cost-performance efficiency, while the M300 targets training and inference for ultra-large multimodal models. Tens of thousands of Kunlun chips are already deployed, and Baidu plans to introduce new chip iterations annually for the next five years.
Meanwhile, Tencent reported quarterly capital expenditures of 20 billion yuan, driven primarily by AI infrastructure, and posted a record 22.82 billion yuan in R&D spending. And Alibaba, according to Barron’s Chinese, has secretly initiated its Qianwen Project, aiming to build a Qwen-model-powered “Qianwen App” to directly challenge ChatGPT and compete for the “AI era endgame.”
Innovation momentum is also rising among emerging model developers. Moonshot AI’s Kimi K2 Thinking, released last week, has garnered substantial attention and may extend the disruptive wave sparked by DeepSeek earlier this year.
Despite global market volatility, China’s AI and GPU sectors remain in rapid-expansion mode. From chip startups racing toward IPOs to tech giants accelerating proprietary chip development and large-model deployment, China’s artificial intelligence industry is entering a period of unprecedented speed and intensity—one increasingly shaped by domestic capability, capital support, and strategic urgency.
Explore more exclusive insights at nextfin.ai.
